· web view2013. 5. 13. · lexical and spatial interference effects in dutch-english bilinguals...
Post on 14-Aug-2021
4 Views
Preview:
TRANSCRIPT
Lexical and spatial interference effects in Dutch-English bilinguals and the need for executive controlWendy van Ginkel, Ton Dijkstra
Supervisors:Ton Dijkstra
Affiliation(s): 1: Donders centre for cognition, Nijmegen, the Netherlands
Corresponding author: van Ginkel, Wendy, Wendyvanginkel@student.ru.nl
Abstract: Inhibitory control originating from within or outside of the bilingual lexicon was assessed by means of two tasks. Thirty Dutch-English bilinguals performed a language decision and a semantic decision task in which they responded selectively to the language or the meaning of Dutch-English direction words (left/right). A spatial interference manipulation reminiscent of the Simon effect was introduced by manipulating stimulus presentation position and response button location. Responses were faster when semantic information matched the location of the response button in both tasks, but the effect of spatial interference arose only in the semantic decision task where it correlated positively with the classic Simon effect. The selective use of lexical information seems to lead to spatial interference effects outside of the bilingual lexicon.
Keywords: Executive control, bilingualism, semantics, language membership
Abstract
Inhibitory control originating from within or outside of the bilingual lexicon was assessed by
means of two tasks. Thirty Dutch-English bilinguals performed a language decision and a
semantic decision task in which they responded selectively to the language or the meaning of
Dutch-English direction words (left/right). A spatial interference manipulation reminiscent of the
Simon effect was introduced by manipulating stimulus presentation position and response button
location. Responses were faster when semantic information matched the location of the response
button in both tasks, but the effect of spatial interference arose only in the semantic decision task
where it correlated positively with the classic Simon effect. The selective use of lexical
information seems to lead to spatial interference effects outside of the bilingual lexicon.
Introduction
2
Many aspects of daily human life require us to selectively attend or respond to specific
aspects of the environment. Intuitive examples range, for instance, from trying to remember
someone’s phone number to the complexity of the selection of scientific content in academic
settings. These situations are governed by a process that is commonly referred to as ‘executive’
control. A widely accepted view of executive control states that three components play a crucial
part: inhibition, updating, and shifting (Miyake et al., 2000). Inhibition refers to the process by
which one can select the appropriate response at the appropriate time, while updating reflects
working memory processes, and shifting concerns more general cognitive flexibility, such as
being able to quickly switch from performing one task to performing another.
A perhaps less intuitive situation in which executive control is thought to play a role is
that of a bilingual speaking two languages. It has been shown that when one is confronted with
object naming in one language, common words in the same and other languages are also
activated to some degree (Green, 1986, 1998; La Heij, 2005). This lexical activation is caused by
a spread of activation through the semantic network to lexical representations of both languages
of the bilingual. For instance, if a Spanish-English bilingual is presented with a picture of a dog,
both names for this animal (‘perro’, and ‘dog’, respectively) will become active to some degree
(Costa, Miozzo, & Caramazza, 1999), just as related semantic representations of DOG such as
TAIL or FUR will also become active (Caramazza, 1997; Levelt, Roelofs, & Meyer, 1999).
Which word is finally selected for production depends on the outcome of competition between
all active words. In other words, the appropriate response at the appropriate time has to be
selected from an array of possible responses. This proposed competition between languages
requires executive control; it has therefore been of great interest to researchers examining the
3
relationship between language and executive control systems in the bilingual, from both
linguistic and non-linguistic perspectives. This line of research has uncovered a so-called
'bilingual advantage' even in cases of non-linguistic executive control. Executive control is
shown to be better developed in bilingual young adults than monolinguals (Bialystok, 2006;
Colzato et al., 2008), to develop earlier in bilingual children than monolingual children
(Bialystok, 2001; Kovacs, 2009), and to be maintained better in older age by bilingual than
monolingual elderly (Bialystok, Craik, Klein, & Viswanathan, 2004; Bialystok, Craik, & Luk,
2008). No general consensus has been reached on what components of control as defined by
Miyake et al. (2000) are the most important for bilingual processing, but the component of
inhibition has been argued to be prominently involved.
The relevance of inhibition as a component of executive control in the bilingual has been
documented by Green (1998), who proposed an inhibitory control model (ICM) in which lexical
representations in the target language compete among each other through inhibition. This
theoretical position has been refined by studies showing that only specific aspects of inhibition as
a component of executive control are enhanced in the bilingual. Particularly relevant to the
present study is the distinction between inhibition of irrelevant information and inhibition of
habitual responses, or “interference suppression” and “response inhibition". Most advantages for
bilinguals are found in the first component of inhibition (Bunge, Dudukovic, Thomason, Vaidya,
& Gabrieli, 2002).
Research into bilingual inhibition has often compared the performance of bilinguals and
monolinguals to determine the nature of the previously mentioned 'bilingual advantage' in
executive control. A task that is commonly used to study inhibitory control is the Simon task,
where participants have to respond to the color of boxes that are presented on the left- or the
4
right-hand side of a computer screen by pressing a button at their right or left hand for a
respective colored box. If a green box on the left-hand side of the screen requires a right button
press, this generates the presence of irrelevant spatial information (the target is on the left, but
the instruction requires a right button press) that causes interference with the response, resulting
in slower reaction times than a congruent button press (pressing the left button if the correct
colored box is presented on the left of the screen). When comparing bilinguals and monolinguals
on this task, bilinguals have been shown to outperform monolinguals (Bialystok, Craik, Klein, &
Viswanathan, 2004). The same pattern has been found in another inhibitory control task, the
flanker arrow task, where participants respond as quickly as possible to a target arrow head that
is flanked by four arrow heads that either point in the same or the opposite direction of the target
arrow that tells participants whether to press the left or right button. Here, relative to
monolinguals, bilinguals also display less interference by flanker arrow heads that are
incongruent with the target arrow head direction (Bialystok, Craik, & Luk, 2008; Emmorey, Luk,
Pyers, & Bialystok, 2008). These studies all employ a design in which monolinguals and
bilinguals are compared to each other on some measure of inhibitory control.
Addressing executive control in a purely bilingual sample poses a challenge, as one loses
the monolingual reference group when implementing a bilingual language task. However,
inhibitory control can be addressed within an individual's bilingual lexicon. Meuter and Allport
(1999) reported a switch cost when bilinguals were asked to name numerals in either their first or
their second language, when these languages were cued unpredictably. When required to switch
from naming the numerals in their first language to naming in their second language, slower
response times were found than when the next numeral was named in the same language. When
switching from their non-dominant second language to their dominant first language, switch
5
costs were found to be even larger. The observed switching asymmetry was explained by the
researchers by proposing that suppressing the dominant language requires more inhibitory
control than suppressing the non-dominant language. As a consequence, it is more difficult to
switch into the dominant language than into the non-dominant language. A similar switch cost
has been observed in language comprehension. The bilinguals' performance on lexical decision
tasks (in which participants decide whether a letter string is a word or not in a specific language),
is slowed by a switch in language of the target word (Thomas & Allport, 2002). The authors
suggested that the locus of this effect should be placed outside of the lexicon, because the
addition of words with a language-specific orthography in one language of the bilingual did not
result in any differences in reaction times. Had the effect originated within the lexicon, words
with a language-specific orthography should have shown faster reaction times, because the
presence of fewer lexical competitors in the second-language pool of the bilingual would
necessitate the recruitment of less inhibitory control. However, no difference for target words
with language-specific orthography was found, suggesting a locus of control outside of the
lexicon.
The precise way in which the languages of a bilingual affect each other during lexical
processing has been more extensively studied using lexical and language decision tasks. In
bilingual lexical decision, participants are asked to specify if a target letter string presented on
the screen is a word or not in a specific language (language-specific lexical decision), or if it is a
word in either of the bilinguals’ languages (generalized lexical decision). In a series of
experiments, Dijkstra, van Jaarsveld, and ten Brinke (1998) found no difference in reaction times
to English-Dutch homographs (words that are written the same in both languages but have a
different meaning in each language) and English control words when there were no Dutch-only
6
words in an English lexical decision task with Dutch-English bilinguals. However, adding Dutch
words to the sample made the task more difficult: Not only did participants have to respond 'no'
to Dutch words that they recognized rapidly, they also had to still respond 'yes' to the interlingual
homographs. The results showed an inhibitory effect on the homographs that was dependent on
their relative frequency in both languages (replicated by Dijkstra, de Bruijn, Schriefers, & ten
Brinke, 2000). This finding provides clear support for the hypothesis of language nonselective
access to the bilingual lexicon, which holds that initially words are activated and considered in
both languages of the bilingual. According to Green (1998), bilinguals might temporarily inhibit
their more dominant language in order to make decisions in their weaker language. Being
required to make decisions about their first and second languages unpredictably might lead
bilinguals to selectively inhibit their first language more than their second language to allow fast
decisions on both languages.
These and other findings in the field of bilingual language recognition have fueled
the development of the Bilingual Interactive Activation Model or BIA model (Dijkstra, & van
Heuven, 1998), which was refined as the BIA+ model (fig. 1) in 2002. In this model, the input
word is parsed on different sublexical and lexical levels. Upon input, candidate words with a
similar sublexical orthography as the input are excited in both languages, whereas those
candidates that lack similar properties as the input are inhibited. This information is fed to other
levels of processing, and subsequently to so-called language nodes, which serve as language tags
for words and receive activation from all lexical representations in a particular language (e.g.,
English or Dutch). Competition between words and languages is resolved through local lateral
inhibition in the word identification system. All information during processing is fed
7
continuously to the task/decision system (task schema), where processing operations to perform
a task are stored. These task schemas are similar to those proposed by Green (1998).
Figure 1. The Bilingual Interactive Activation + (BIA+) model (Dijkstra, & van Heuven, 2002).
The current study
In the BIA+ model a distinction is made between language nodes, which denote
language membership, and semantics. These two properties of words are hypothesized to be
separate aspects of the lexical representation of words and do not directly interact. The current
study employed this aspect of the model to investigate the relationship between bilingualism and
inhibitory control within and outside the lexicon in a purely bilingual participant population. To
this end, Dutch-English direction words were used as target stimuli in a language decision task
and in a semantic decision task performed by Dutch (L1) - English (L2) bilinguals. In both tasks,
the English and Dutch words 'left', 'right', 'links' (the Dutch equivalent of 'left'), and 'rechts'
(Dutch for 'right') were presented on either the left- or right-hand side of a computer screen.
8
From a semantic perspective, direction words are 'special' in that they have a meaning that can be
related to nonverbal aspects of the task, such as screen position and button location. It seems
likely that the direction word itself, its position on the computer screen, and its associated
response button, will all involve the activation of semantic spatial information (referring to 'left'
or 'right' in the spatial domain). By combining different direction words with varying screen
positions and linking them to different button locations, interference effects between lexical and
non-lexical spatial representations can potentially be elicited. This allows a study of the
interactions between lexical and spatial information in the language decision and semantic
decision tasks.
In the language decision task, Dutch-English bilinguals categorized four direction words
according to language membership by pressing either a left or a right button for a Dutch or an
English categorization, depending on the button configuration at hand. The four target words
were 'left' and its Dutch equivalent 'links', and 'right' and its Dutch equivalent 'rechts'. In this
task, participants had to ignore both the position on the screen where the word appeared and its
meaning ('left' or 'right'). Their response had to be based instead on the language of the item
(Dutch or English) as it was associated with one of two response buttons.
In the semantic decision task, participants were asked to either press the button
corresponding to the meaning of a word (e.g., pressing the left button if the target was 'left') or to
press the button opposite to the meaning of the target (e.g., pressing the left button if the target
was 'right'). Here the position of the word on the screen and its language membership were
irrelevant for the required response.
To summarize, the target words in both the language decision and semantic decision tasks
were exactly the same. Furthermore, in both tasks spatial aspects that were irrelevant to the
9
specific instruction could be activated, depending on the stimulus position on the screen or on the
response hand (left or right). Thus, the two tasks in our study had an intended similarity to the
Simon task, where irrelevant spatial information plays an important role: In both tasks, stimuli
are presented on the side of the screen that was either congruent or incongruent with the location
of the required response button. To directly compare the classic Simon effect to the effect of an
incongruency between screen position and button location in our experimental tasks, we included
the Simon task in our experiment. We hypothesized that a Simon-like interference effect should
only occur in our semantic decision task, where participants were required to selectively respond
to the semantics of the target (rather than language membership). In the Predictions section
below, this specific prediction and other predictions will be discussed in more detail. Next to
language decision, semantic decision, and a Simon task, participants also filled out an extensive
language background questionnaire with queries regarding their skill on different dimensions of
their L2, their own and their friends' frequency of use of L2 words in L1 conversation (language
switches), and the number of other languages they were familiar with next to their L1 and L2. As
bilinguals have been shown to have smaller vocabularies than monolinguals (Perani et al., 2003;
Portocarrero, Burright, & Donovick, 2007), and vocabulary size has been shown to be related to
rapid retrieval of lexical information (Beck, Perfetti, & McKeown, 1982; Hedden,
Lautenschlager, & Park, 2005), participants also performed a vocabulary task to take into
account the L2 vocabulary size of individual participants.
Predictions
As participants would be asked to selectively attend to language membership information
in the language decision task, we expected language to be a main factor in this task, resulting in
10
faster responses on targets in the dominant L1, Dutch, than on targets in the weaker L2, English.
Bearing in mind the proposition by Green (1998) that bilinguals might be able to selectively
inhibit a stronger language to perform on one of their weaker languages, this language effect
could be rather small. Furthermore, although participants were asked to make a decision based
solely on language membership, the semantics of the target word might still receive some degree
of processing and it might be able to interfere with the language decision through incongruency
with the required response button. In addition, a main effect of button location was predicted to
arise as the participant group for the experiment was right-handed. For this reason, response
times on the right button should be faster than response times on the left button.
For the semantic decision task, we also expected this effect of button location due to
handedness. Furthermore, as participants were required to selectively use the semantic
information for their button press, we predicted an interaction between the semantics of the word
and the location of the button. Specifically, if the target word was in the 'left' grouping of words
(e.g., the target words 'links' and 'left'), we expected reaction times on the right button to be
slowed due to an incongruency with the meaning of the word. Most importantly, we predicted an
interaction of screen position and button location in the semantic decision task. Due to the
selective emphasis on the semantics of the word required to perform this task, meaning retrieval
of the target word might give rise to interference effects outside of the lexicon, at the level of
task execution. Direction words should activate their inherent semantic properties in this task,
denoting direction, as participants are required to make decisions based on this property.
However, the direction that a target denotes could be interfered with by other aspects of the task,
in particular a representation of screen position and of button location. For example, words with
the semantic information of ‘left’ (to be referred to as S-left from this point onwards), might
11
cause a greater spatial interference effect in cases where both the screen position and button
location are to the right side. In this case, two factors point to one direction whilst the semantics
indicates another direction. In the case of S-left presented on the left of the screen and requiring a
right button press, there is only one factor that opposes the direction inherent to the semantics of
the target word (see table 1). This might give rise to a three-way interaction between the
semantics of the word, the screen position that the word is presented at, and the location of the
correct response button.
Table 1. Congruency of words semantics with screen position and button location.
Screen position Button location Congruency
Semantics S-Left Left Left Congruent with all other factors
Left Right Incongruent with button location
Right Left Incongruent with screen position
Right Right Incongruent with all other factors
S-Right Left Left Incongruent with all other factors
Left Right Incongruent with screen position
Right Left Incongruent with button location
Right Right Congruent with all other factors
The specific predictions for the two experimental tasks are visually represented in figure
2 below, where the most relevant effect is marked by the large bold arrow, showing how the
requirement to attend to word semantics is hypothesized to relate to screen position and button
location, and subsequently to fuel an interaction between these two. The left of the model refers
to predictions regarding the language decision task (through language nodes) and the right of the
model refers to predictions regarding the semantic decision task (through meaning information).
12
Figure 2. Visualization of the language decision and semantic decision task.
To be able to relate the predicted spatial interference effect in the semantic decision task
to a more classic demonstration of this effect, a Simon task was also administered to the
participants. This task should, of course, result in the classic Simon effect: Stimuli presented at a
screen position that is incongruent with the location of the required response button (e.g., a
stimulus presented on the left of the screen that requires a right button press) should be
responded to slower than stimuli presented at a screen position that is congruent with the location
of the required button, showing the spatial interference effect. Because we hypothesized that the
specific demand to attend to semantic information in the semantic decision task should lead to a
Simon-like interference effect of spatial information (e.g., screen position and response button
location), we hypothesized that responses in these two tasks should correlate positively, with
larger interference effects in one task related to larger interference effects in the other task.
13
In relation to the vocabulary task and the language background measures, we
hypothesized a possible negative relationship of vocabulary size and the speed of retrieval of
language membership and semantic information, where a larger L2 vocabulary should be paired
with faster reaction times on the experimental measures. In relation to the language background
measure, we expected several of the measures to relate to each other. For example, participants
who reported a younger age of acquisition of their L2 were also expected to report more years of
experience with their L2. Participants with a more positive attitude towards the use of L2 words
in L1 conversation were also expected to use L2 words in L1 conversation more often
themselves.
Method
Participants
Participants were Dutch-English bilingual students of the Radboud University in
Nijmegen, the Netherlands (9 men and 21women, 30 in total), aged 18 to 27 years old (mean age
= 19.4, SD = 2.0). All participants were right-handed native speakers of Dutch with normal or
corrected to normal vision, and reported no language deficits such as dyslexia. On average, they
had 7.7 (SD = 1.1) years of educational experience with English (age of acquisition on average:
10.0, SD =1.5). At the end of the experimental session, participants completed a language
background questionnaire and vocabulary test of English. The results of all tests and further
participant characteristics are summarized in table 6 in the results section. All participants were
granted course credit for participation in the experiment.
Stimulus Materials and Design
14
The experiment consisted of several parts, the most important the language decision and
semantic decision task. Next to these tasks, participants performed a Simon task, a vocabulary
task, and filled out a language background questionnaire.
Regarding the language decision and semantic decision tasks, the exact same stimulus set
was used for both, making the mental task that the participant was performing the only difference
between the two. Stimuli in both tasks consisted of four words in two languages: two English
direction words “left” and “right” and two equivalent Dutch direction words “links” and
“rechts”. These words were presented on the right- or left-hand side of the screen and required
responses on a right- or left-hand button, mimicking the effect of irrelevant spatial information as
studied in the Simon task.
In the language decision task, participants were required to press one of two buttons to
indicate whether the word on the screen was either Dutch or English. Participants performed the
language decision task twice: with both possible button configurations (once with Dutch on the
left and English on the right button, and once with English on the left and Dutch on the right
button). The order of block presentation was counter-balanced across participants (see
Randomization and counterbalancing). Participants also performed a semantic decision task with
two possible button configurations, where they had to either press the button that corresponded
to the word on the screen (pressing the left button if the word on the screen meant left in either
language), or they had to press the button opposite to the word on the screen (pressing the right
button if the word on the screen meant left in either language). In this way, the only difference
between the two tasks was the mental task that the participants were performing: either deciding
on the language membership of a relevant word, or on the semantic properties of the word.
15
Next to the two experimental tasks, participants performed a Simon task with the purpose
of correlating performance on the semantic decision task with this more classic measure of
inhibitory control. This allows a check of the assumption that the two measures are actually
similar in the respect that the irrelevant spatial dimension of the target stimulus should interfere
with the response, whereas in the language decision task this should not be the case.
For the subsequent analyses, the reaction times of the button presses to the words on the
screen and whether the button press was correct were used.
Apparatus and Procedure
The experiment was performed on a HP desktop computer connected to a 21 inch screen,
and participants were seated at an approximate distance of 60 cm to the computer screen.
Upon arrival, participants were randomly assigned to one of the possible task sequences
in the experiment that are discussed below, and then performed the experimental task. Secondly,
they performed the Simon task. Next, they were asked to fill out a language background
questionnaire and perform an English vocabulary test. Both questionnaires and experimental
tasks are discussed in more detail next.
The experiment
Randomization and counterbalancing. Participants performed a language decision and
semantic decision task interchangeably, in a counter-balanced order. For example, a participant
might have started with a language decision task, then performed a semantic decision task, then
language decision again, and then semantic decision again. Note that all possible button
configurations were presented to the participant: one language decision task with English on the
left and Dutch on the right button, and one with English on the right and Dutch on the left button.
16
The same principle was applied to the semantic decision task: Participants performed this task
once by pushing the left button if the word on the screen meant left and pushing the right button
if it meant right in either language, and once by pushing the right button if the word meant left
and pushing the left button if it meant right in either language.
Furthermore, four different stimulus lists were created for every single one of the four
task configurations, resulting in a total of 16 different pseudo-random stimulus sequences. These
lists were then counter-balanced across participants again. For example, a participant might
receive list 1 for the language decision with English on the left button, list 2 for the semantic
decision with ‘left’ on the left button, list 3 for language decision with English on the right
button, and finally list 4 for semantic decision with the reverse button configuration (pressing the
left button when the word means ‘right’). Internally the order within the lists was the same,
meaning that list 1 for the language decision variant with English on the left button was the same
in terms of the order of position-button congruent and incongruent trials as list 1for the language
decision variant with English on the right button. Participants never received both of those lists:
each list for each task had a different number, ensuring that they never had the same order of
congruent/incongruent trials across the different tasks.
Each task was preceded by a short practice session consisting of 16 trials: eight congruent
trials (four on the left and four on the right button) and eight incongruent trials (four on the left
and four on the right button again). After the practice session, participants pressed a button to
start the actual task.
Language decision task. In the language decision task, participants were presented with
four direction words in two languages: “left” and “right” in English and their Dutch equivalents
“links” and “rechts”. Before each trial, a fixation cross was presented in the centre of the screen,
17
and it remained there for 1000 ms. Then, one of the four possible words was presented on either
the left- or the right-hand side of the computer screen. Participants were required to press a
button to indicate whether the word on the screen was English or Dutch, and to do so as fast and
yet as accurately as possible. They were given a 1500 ms time window to respond to a word,
before the next word was presented. Each participant was given two versions of the task: in one
version of the task, the response button for English was on the left and the response button for
Dutch was on the right, and in another version these button positions were reversed. These tasks
are referred to as A1 and A2 in figure 3.
Figure 3. Visualization of the language decision task. Picture a depicts version A1 of the language decision task, where
an English response is given on the left button and a Dutch response is given on the right button. Picture b depicts
version A2 of the language decision task, where the response buttons are reversed.
Words were presented in two blocks of 48 trials, creating a total of 96 stimuli for each
version of the language decision task, totaling 192 stimuli across the two versions of the task.
After the first block of 48 trials, participants were given the opportunity to have a break. To
continue the task (and to start the task after instruction screens), participants pressed the middle
button on their response box, so the following response button (on the first experimental trial)
18
+English
Dutch
left1000 ms
1500 ms
+Dutch
English
left1000 ms
1500 ms
a bVersion A1
Version A2
could not coincide with the button they had used to continue from the instruction screen, to
exclude any motor effects from this response on the first experimental stimulus.
Semantic decision task. In this task, the same words were presented as in the language
decision task to optimize the comparison between the two tasks. However, instead of responding
to the language in which the word was presented, participants had to respond to the semantics of
the direction words. In one version of the task, participants were required to press the button
congruent to the semantics of the target direction word. So, if the word 'left' was presented on the
screen, they had to press the left button, irrespective of the position that the target was presented
at on the screen. In the second version of the task, participants were required to do the opposite
of the target word: if one of the direction words “left” or “links” was presented, participants had
to press the button at the position of their right hand on the button box. If the word “right” or
“rechts” was presented they had to press the button at their left hand. In this way, they were
required to ignore the language information and focus solely on the semantics of the direction
words, whereas in the language decision task they were required to ignore the semantics of the
words and focus solely on the language. Again, they were required to respond as quickly but as
accurately as possible. A total of 192 stimuli were presented, equally divided over four subtasks:
two where the participant had to press the left button and two where the participant had to push
the right button. The semantic decision task played into the difficulty of ignoring spatial
information in a word: participants were expected to show more difficulty pressing a button on
the left if the word indicated a right direction. Again, the reaction times and accuracy of
responses were collected. These tasks are referred to as B1 and B2 in figure 4 below.
19
Figure 4. Visualization of the semantic decision task. Picture a depicts version B1 of the semantic decision task, where
a 'left' response is given on the left button and a 'right' response is given on the right button. Picture b depicts version
B2 of the semantic decision task, where the response buttons are reversed.
Simon task. Targets in the Simon task consisted of green and blue boxes that were
presented on either the right or the left of the computer screen. Participants responded to the
color of the boxes, where a green box required a left hand response and a blue box required a
right hand response. Each trial began with a fixation cross (+) that lasted 1000 ms, before the
trial word was presented on either the left- or the right-hand side of the screen. The target stayed
on the screen for approximately 1000 ms, and participants had a total time window of 1500 ms to
respond to the target by the addition of a 500 ms blank screen that followed each trial. During the
blank screen, participants could still provide their response even though the target had
disappeared. The next trial was then again signaled by the appearance of the fixation cross, etc.
The task was split into two blocks of 60 trials, creating a total of 120 trials divided over four
options: two options where the stimulus position on the screen was congruent with that of the
location of the response button (a green box appearing on the left of the screen or a blue box
appearing on the right of the screen) and two options where these locations were incongruent (a
20
+left
right
left1000 ms
1500 ms
Version B1
a
+right
left
left1000 ms
1500 ms
Version B2
b
green box appearing on the right of the screen or a blue box appearing on the left of the screen).
Reaction times from the point of presentation of the target stimulus to the actual button press and
accuracy scores were recorded.
Language background questionnaire. This questionnaire included items on the
participants’ language use in daily life, their language background, their attitude towards
language-switching, and a self-report of their fluency in their second language (English) in
reading, writing, speaking, and listening (on a scale of 1 to 7, 7 being ‘the same as native
language’). Items on language background included self-reported knowledge of languages other
than those included in the experiment, as well as age of acquisition of English. Questions
concerning the participant’s attitude towards language switching addressed issues of L2 words
being used in L1 by other people (ranging from 0 -it annoys me- to 5 -I think it’s fun-) and how
often the participant uses L2 words in L1 themselves (on a scale from 0 -never- to 5 -often-).
Furthermore, the participants were asked reasons for their own use of L2 words in their L1
speech.
Vocabulary test. To assess participant proficiency in English, an adjusted and
computerized version of the off-line vocabulary test by Meara and Jones (1990) was used.
Participants were presented with words and non-words, and they were asked to determine
whether the word they were presented with was either an existing English word by clicking a
happy face, or whether it was not an English word by clicking a sad face presented in the
program. The percentage correctly classified words was calculated and used in the analyses as an
indication of L2 vocabulary size.
21
Results
Participants were excluded from analysis if their response accuracy dropped significantly
below 50% on any of the tasks or if they performed a task in an opposite manner to the one
instructed (pressing left for Dutch if the instruction was to press right). According to the first
criterion, 6 participants were excluded. A further 4 were excluded on the second. For the
language decision task, the overall accuracy rate of the participants included in the analysis was
88.27% with a standard deviation of 10.5, whereas for the semantic decision the accuracy rate
was 88.28% with a standard deviation of 21.3. Considering the entire dataset, there was no
significant difference in accuracy between the two experimental tasks. Accuracy rates for all
participants are summarized in table 2 in appendix a.
To ensure that the data in the analysis were an accurate reflection of participants’
responses, participants were also excluded if a large portion of their response times were
centered at the high end of the response time limit. Participants had a 1500 ms window to
respond to the target word on the screen, and no reaction times above this time were recorded. If
a participant showed reaction times higher than this limit for 25% of the total trials in a specific
condition of a specific task, this participant was also excluded from analysis as it would not be
clear what their actual response times would have been. To trim outliers, any response time 2
standard deviations below or above the average response time for a person were also excluded
from the analyses, just as incorrect trials.
Main effects
Main effects were examined by concatenating data of the three factors that were not the
factor of interest for the current main effect analysis over the data of the factor of interest. For
22
example, in order to examine the main effect of the factor Position on the screen (left/right), two
different means (Ms) were calculated from the reaction times (RTs) for both the language
decision and the semantic decision task: one mean for instances where the target word appeared
on the left of the screen, and one mean for instances where the target appeared on the right of the
screen. Differences in other factors were disregarded by including both dimensions of those
factors in the two means for the main effect of the relevant factor, thus only considering
variations in the relevant factor. The data are summarized in table 3 below.
Table 3. Means and standard deviations (sd) on single factor data for Position on the screen (left or right), meaning of the Word
(grouping of ‘left’ and ‘links’ vs. ‘right’ and ‘rechts’), response Button (left or right) and target Language (Dutch and English).
Separate means are shown for the relevant factor, the dimensions of the three other factors are considered together within these
separate means.
Position Button
left right overall left right overall
Language decision
Task1 Mean (sd) 707 (82) 696 (77) 702 (80) 702 (84) 701 (76) 702 (80)
Accuracy (sd) 89.3 (9.5) 90.0 (8.2) 89.7 (8.9) 90.1 (9.1) 89.3 (8.3) 89.7 (8.7)
Semantic decision
Task2 Mean (sd) 644 (83) 630 (80) 637 (82) 641 (86) 634 (77) 638 (82)
Accuracy (sd) 94.5 (3.1) 93.5 (3.7) 94.0 (3.4) 94.2 (3.4) 93.8 (3.1) 94.0 (3.3)
Word Language
left right overall Dutch English overall
Language decision
Task1 Mean (sd) 699 (81) 705 (79) 702 (80) 705 (79) 798 (81) 752 (80)
Accuracy (sd) 90.1 (8.6) 89.2 (9.2) 90.0 (8.9) 89.8 (9.2) 89.6 (8.5) 89.7 (8.9)
Semantic decision
Task2 Mean (sd) 636 (82) 638 (80) 637 (81) 630 (82) 645 (81) 638 (82)
Accuracy (sd) 93.9 (3.4) 94.1 (3.7) 94.0 (3.6) 94.8 (2.8) 93.2 (3.9) 94.0 (3.4)
Effects were investigated using multiple repeated measures ANOVAs with both the
relevant factor and the factor Task (Language decision (Task1) / Semantic decision (Task2)) as
within subject factors. Accuracy data were investigated in the same manner. Each subsection
23
first discusses the accuracy analysis of the relevant factor, and then the RT analysis.
Furthermore, the tasks are first compared on accuracy rates and RTs, before their separate
analyses are discussed.
Factor: Position. The accuracy analysis for the factors Position (left/right on the screen)
and Task (language decision / semantic decision) showed higher accuracy in the semantic
decision than in the language decision task (F (1, 29) = 9.893, p < .005, eta²= .254). The overall
RT analysis showed faster RTs in the semantic decision than in the language decision task (F
(1,29) = 59.019, p <.0001, eta² = .671). There was an overall main effect of Position (F (1,29) =
14.524, p <.001, eta² = .334), where RTs in response to targets appearing on the right of the
screen were smaller than RTs for targets appearing on the left of the screen. This effect of
Position was found both in the language decision task (F (1,29) = 5.497, p <.027, eta² = .159) and
the semantic decision task (F (1,29) = 12.006, p <.002, eta² = .293).
Factor: Button. The accuracy analysis for the factors Button (left/right button) and Task
showed higher accuracy in the semantic decision than in the language decision task (F (1,29) =
9.893, p <.005 , eta² = .254). The overall RT analysis showed faster RTs in the semantic decision
than in the language decision task (F (1,29) = 58.379, p <.0001, eta² = .668), but no main effect
of Button was found in either of the tasks.
Factor: Word. The accuracy analysis for the factors Word (S-left/S-right) and Task again
revealed higher accuracy in the semantic decision than in the language decision task (F (1,29) =
9.893, p <.005, eta² = .254). The RT analysis showed faster RTs in the semantic decision task (F
(1,29) = 58.575, p <.0001, eta² = .669) than in the language decision task. No main effect of
Word was found in either of the experimental tasks.
24
Factor: Language. The accuracy analysis for the factors Language (Dutch/English) and
Task showed higher accuracy in the semantic decision than in the language decision task (F
(1,29) = 9.893, p <.005, eta² = .254). The RT analysis did not only show faster RTs in the
semantic decision than in the language decision task (F (1,29) = 58.429, p <.0001, eta² = .668),
but also showed a significant Task*Language interaction effect (F (1,29) = 14.502, p <.002, eta²
= .333). Separate analyses of the tasks revealed that the origin of this interaction stemmed from
the fact that there was no main effect of Language in the language decision task, but there was a
main Language effect in the semantic decision task (F (1,29) = 16.542, p <.0001, eta² = .363),
with faster responses to Dutch than to English words.
Interactions
Relevant interaction effects between factors were studied separately. Again, the data for
repeated measures ANOVAs consisted of variables that created a distinction within and between
the relevant factors, and that included both dimensions on the irrelevant factors. For example, to
study the interaction between the factors Position and Button (in this study one condition of this
interaction in the semantic decision task functioned as the hypothesized equivalent of the classic
Simon effect, see section Comparing the Simon and experimental tasks), a repeated measures
ANOVA was conducted with within subject factors Task (language decision/semantic decision),
Position (left/right (on the screen), and Button (left/right button). Eight variables were defined by
recoding the available data, for each task per participant: (1) a mean for position left in
combination with button left, (2) a mean for position left in combination with button right, (3) a
mean for position right in combination with button left, (4) a mean for position right in
combination with button right. These means were entered into the analysis and examined in a
25
similar fashion for all relevant interactions (see table 4 for a summary of the data and figure 5 for
a visual representation of the interactions).
Table 4. Means and standard deviations (sd) for all relevant two-way interactions. Mean accuracies (and standard
deviations) are listed as percentages of correct trials for a specific type of trial.
Interaction: Position*Button Interaction: Position*Button
Language Button Semantic Button
Task 1 left right overall Task 2 left right overall
Position left Mean (sd) 710 (92) 704 (77) 707 (85) Position left Mean (sd) 637 (91) 652 (78) 645 (85)
Accuracy (sd) 90.0 (10.2) 88.6 (9.9) 89.3 (10.1) Accuracy (sd) 94.8 (4.0) 94.2 (3.1)94.5 (3.6)
right Mean (sd) 694 (81) 698 (78) 696 (80) right Mean (sd) 644 (85) 615 (80) 630 (83)
Accuracy (sd) 90.1 (9.1) 89.9 (8.0) 90.0 (8.6) Accuracy (sd) 93.7 (4.7) 93.3 (4.2)93.5 (4.5)
overall Mean (sd) 702 (87) 701 (78) overall Mean (sd) 641 (88) 634 (79)
Accuracy (sd) 90.1 (9.7) 89.3 (9.0) Accuracy (sd) 93.8 (4.4) 93.8 (3.7)
Interaction: Word*Button Interaction: Word*Button
Language Button Semantic Button
Task 1 left right overall Task 2 left right overall
Word left Mean (sd) 682 (93) 717 (87) 700 (90) Word left Mean (sd) 602 (91) 672 (90) 637 (91)
Accuracy (sd) 93.4 (7.8) 86.8 (10.5) 90.1 (9.2) Accuracy (sd) 95.3 (4.4) 92.5 (4.7)93.9 (4.6)
right Mean (sd) 725 (92) 687 (71) 706 (82) right Mean (sd) 681 (99) 597 (86) 639 (93)
Accuracy (sd) 86.8 (10.5) 91.7 (8.2) 89.3 (9.4) Accuracy (sd) 93.1 (4.4) 95.0 (4.3)94.1 (4.4)
overall Mean (sd) 704 (93) 702 (79) overall Mean (sd) 642 (95) 635 (88)
Accuracy (sd) 90.1 (9.2) 89.3 (9.4) Accuracy (sd) 94.2 (4.4) 92.3 (4.5)
Interaction: Language*Button Interaction: Language*Button
Language Button Semantic Button
Task 1 left right overall Task 2 left right overall
Language Dutch Mean (sd) 708 (97) 706 (76) 707 (87)Language Dutch Mean (sd) 629 (88) 630 (78) 630 (83)
Accuracy (sd) 89.2 (16.3) 90.3 (5.2) 89.8 (10.8) Accuracy (sd) 95.2 (4.0) 94.3 (3.9)94.8 (4.0)
English Mean (sd) 699 (84) 695 (86) 697 (85)
English Mean (sd) 653 (89) 637 (77) 645 (83)
Accuracy (sd) 91.0 (5.3) 89.6 (8.5) 90.3 (6.9) Accuracy (sd) 93.3 (4.8) 93.2 (3.9)93.3 (4.4)
overall Mean (sd) 704 (91) 701 (81) overall Mean (sd) 641 (89) 634 (78)
Accuracy (sd) 90.1 (10.8) 90.0 (6.9) Accuracy (sd) 94.3 (4.4) 93.8 (3.9)
26
Language decision task Semantic decision task
27
Figure 5. Visualization of the interactions with separate line graphs for each task, including the factors Position
(PosLeft/PosRight), Button (ButLeft/ButRight), Word (WordLeft/WordRight) and Language (Dutch/English).
Interaction: Position*Button. The accuracy analysis for the interaction between the
factors Task, Position and Button showed that participants made fewer errors in the semantic
decision than in the language decision task (F (1,29) = 9.893, p <.005, eta² = .254). The RT
analysis for the Position*Button interaction showed that RTs in the semantic decision task were
also faster than those in the language decision task for this data configuration (F (1,29) = 58.501,
p <.0001, eta² = .669). Responses were faster when there was congruency between the position
of the target on the screen and the response button, as evident by a Position*Button interaction (F
(1,29) = 5.905, p <.023, eta² = .169). However, the Task*Position*Button interaction showed
that this interaction was not the same in both tasks (F (1,29) = 25.147, p <.0001, eta² = .464).
Separate analyses of the experimental tasks showed a Position*Button interaction only in the
semantic decision task (F (1,29) = 21.414, p <.0001, eta² = .425), but not in the language
decision task. This finding was in accordance with the hypothesis that the semantic decision task
should to some degree cause interference between target and button position that mimics the
interference in the Simon task. The RT analyses of both experimental tasks showed that
28
responses to targets appearing on the right of the screen were faster, but this effect was slightly
larger in the semantic decision (F (1,29) = 12.472, p <.002, eta² = .300) than in the language
decision task (F (1,29) = 5.409, p <.028, eta² = .157).
Interaction: Word*Button. The accuracy analysis for the interaction between the factors
Task, Word and Button again showed fewer errors in the semantic decision task (F (1,29) =
9.916, p <.005, eta² = .255), but also yielded a significant Word*Button interaction (F (1,29) =
22.421, p <.0001, eta² = .436), where fewer errors were made when there was congruency
between the semantics of the word and the response button (e.g., the word ‘left’ coupled to a left
response) and this interaction differed slightly between both tasks (F (1,29) = 6.837, p <.015, eta²
= .191). The RT analysis showed the same patterns as the accuracy analysis: a significant
difference between the tasks with faster RTs on semantic decisions (F (1,29) = 58.512, p <.0001,
eta² = .669), as well as a Word*Button interaction (F (1,29) = 39.638, p <.0001, eta² = .577).
Furthermore, the interaction effect between the factors Word and Button was shown to differ
across the tasks (F (1,29) = 9.072, p <.006, eta² = .238). Whilst both the language decision task
(F (1,29) =29.732, p <.0001, eta² = .506) and the semantic decision task (F (1,29) = 28.737, p
<.0001, eta² = .498) show a Word*Button interaction of about the same magnitude, there is a
slight difference in the nature of the interaction: in the language decision task RTs for responses
to targets in the 'left' semantic category that require a left button press are faster than targets in
the 'right' category requiring a right button press, whilst in the semantic decision task this effect
is reversed with faster RTs for a 'right' category target on the right button. As there was a main
effect of Language for the semantic decision task, we looked at the Language*Word interaction
and the Language*Word*Button interaction in both tasks to investigate whether the effect of
29
semantics was dependent on the language of the word. Both of these interaction effects were not
found in either task.
Interaction: Language*Button. The accuracy analysis for the factors Task, Language and
Button showed fewer errors in the semantic decision than in the language decision task (F (1,29)
= 11.959, p <.003, eta² = .292), but no other differences in accuracy were found. The RT
analysis showed faster RTs in the semantic decision than the language decision task (F (1,29) =
57.675, p <.0001, eta² = .665) and a significant Task*Language interaction (F (1,29) = 15.841, p
<.0001, eta² = .353). As discussed in the main factor analyses, this interaction stemmed from the
absence of a main effect of Language in the language decision task, whereas this effect was
present in the semantic decision task (F (1,29) = 16.915, p <.0001, eta² = .368). Furthermore, the
Language*Button interaction also was shown to have a small effect in the semantic decision task
(F (1,29) = 4.723, p <.05, eta² = .140).
Higher order interactions. Of specific interest to the present study were interactions
between the relevant factor for the task at hand and the Position and Button factors. No higher
order interaction between the factors Language, Position and Button was found for either the
language decision task (F (1,29) = .212, p < .65), or the semantic decision task (F (1,29) = .132, p
<.720). Also, Word*Position*Button interaction was found in either the language decision task
(F (1,29) = 2.830, p <.104) or the semantic decision task (F (1,29) = 1.077, p <.309).
Simon task
All reaction times 2 standard deviations above and below the means for a certain
participant were excluded from analysis. All participants that were included in the analyses of
the language decision and the semantic decision task were included in the Simon task analysis.
30
For analysis of the Simon task, a repeated measures ANOVA with factors Position
(left/right on the screen) and Button (left/right) was conducted on both accuracy and RT data.
Means corresponding to this analysis are shown in table 5. The accuracy analysis did not show
any differences in accuracy for any of the single factors or for the Position*Button interaction,
suggesting no differences in accuracy rates for any of the analyses. In the RT analysis, a main
effect was found for the factor Button (F (1, 29) = 7.295, p <.011, eta²=.201), where responses on
the right button were significantly faster than responses on the left button. As all participants
were right-handed, this result is not surprising. More interesting is the interaction between the
factors Position and Button, which was also shown to be significant (F (1, 29) = 56.644, p
<.0001, eta²=.661). This is a demonstration of the classic Simon effect, where responses on the
button corresponding to the side of the screen that the stimulus is presented on are faster than
responses to stimuli on the opposite side of the screen (fig. 6). For example, a response to a
stimulus on the left of the screen is faster when the participant is required to press the left button
than when the participant is required to press the right (incongruent with the stimulus position)
button.
Table 5. Means (and standard deviations) corresponding to reaction times and accuracy percentages in the Simon task. Table (a)
displays single factor data and table (b) displays interaction data.
(a) Position Button
left right overall left right overall
RT (sd) 450 (80) 439 (78) 446 (79) 447 (79) 442 (79) 445 (79)
Accuracy (sd) 98.3 (2.6) 97.4 (3.2) 97.9 (2.9) 97.6 (3.3) 98.1 (2.4) 97.9 (2.9)
(b) Button
left right overall
Position left RT (sd) 439 (82) 455 (79) 447 (81)
Accuracy (sd) 98.6 (3.0) 98.1 (2.9) 98.4 (3.0)
right RT (sd) 462 (80) 424 (81) 253 (81)
Accuracy (sd) 98.1 (2.9) 98.6 (2.3) 98.4 (2.6)
31
overall RT (sd) 451 (81) 440 (80)
Accuracy (sd) 98.4 (3.0) 98.4 (2.6)
Figure 6. Graph of the Position*Button
interaction in the Simon task.
Comparing the Simon and experimental tasks
To compare the data from the two experimental tasks with that of the Simon task, a
correlation analysis was performed on the difference scores of incongruent and congruent trials
in all tasks on the factors Position and Button. All trials where a target stimulus was presented on
the left of the screen and correct button response was a left button response and vice versa were
defined as congruent trials, and all trials where a target was presented on the left of the screen
but the correct button response was a right button response and vice versa were defined as
incongruent trials. Difference scores were used in the analysis and were computed by subtracting
the mean RTs of congruent trials from those of the incongruent trials for all participants in all
tasks.
The correlation analysis showed no significant correlation between the language decision
task and the Simon task, or between the language decision and semantic decision tasks.
However, there was a significant correlation between the difference scores on the Simon and the
semantic decision task (r = .400, p <.029). This correlation confirms the hypothesis that the
semantic decision task elicited a Position*Button interference effect reminiscent of that in the
Simon task, because participants with a higher Position*Button interference effect in the
semantic decision task also show an elevated Position*Button interference effect in the Simon
task.
32
Analyzing the language background questionnaire data
In order to create a clear picture of the language background of the participants, all
participants filled out a language background questionnaire with questions on their knowledge of
languages and specific attitude and use of English in the Dutch language. Mean scores on these
measures are summarized in table 6 below. Listed are the participants’ age, score on the
vocabulary measure, the number of languages known totally to the participant, the age at which
they started learning their L2 (English), how many years of education they enjoyed in their L2,
their self-reported speaking, writing, listening, and reading skill in their L2 (on a scale of 1 -I
don’t write/speak/listen/read this language- to 7 -like my mother tongue-) and finally their
opinion of the use of L2 words in L1 by others (on a scale of 1 -it annoys me- to 5 -I like it-) and
their own use of L2 words in L1 (on a scale of 1 -never- to 5 -very often-). In the following
subsection correlations between background measures and tasks will be discussed.
Table 6. Means (and standard deviations) on the language background questionnaire data.
Age VocabularyNumber of languages known
Age ofAcquisitionL2
Education (in years)
Speaking Writing Listening Reading Attitude L2 in L1
Use L2 in L1
Friends use of L2 in L1
Mean (sd)
19.4(2.0)
75.6(9.3)
4.1(0.8)
10.0(1.5)
7.7(1.1)
5.5(0.7)
5.7(0.9)
6.0(0.7)
6.2(0.8)
3.3(1.0)
3.3(1.1)
3.3(0.7)
T-tests were conducted on the self-reported skill scores to see how participants rated their
skills against each other. In this analysis, it became apparent that participants rated their speaking
skill considerably lower than their listening skill (t = -4.267, p < .0001) and their reading skill (t
= -4.583, p <.0001), but not than their writing skill. Also, the overall rating of the participants’
writing skill was lower than that of their listening skill (t = -2.921, p <.008), and their reading
skill (t = -5.288, p <.0001). This analysis shows that in general, participants scored their
33
comprehension (listening and reading) skills as better developed than their production (writing
and speaking) skills.
Correlations between language background measure scores. Relationships between
language background questionnaire items were investigated by entering the language background
data into a correlation analysis. This analysis revealed that participants who reported knowing a
higher number of languages also enjoyed more years of education in their L2 than participants
who reported knowing fewer languages (r = .381, p <.042), and they also showed a higher self-
reported speaking skill score (r = .537, p <.003). Participants who reported they had starting
learning their L2 at an earlier age (age of acquisition score) also reported having enjoyed more
years of education in their L2 (r = -.556, p = <.003). Participants who reported a more positive
attitude towards the use of L2 words in L1 conversation by other people reported using L2 words
in L1 more often themselves as well (r = .596, p <.002), and reported that their friends used L2
words in L1 more often (r = .405, p<.027) than participants who listed a less positive attitude
towards the use of L2 words in L1 conversation. Finally, participants who reported that their
friends used L2 words in L1 conversation more often rated their own speaking skill as higher
than participants who reported that their friends used L2 words in L1 conversation less often (r
= .459, p <.012) and reported using L2 words more often in conversation themselves (r = .782, p
<.0001). Participants who reported using L2 words in L1 conversation more often themselves
also showed the same higher self-reported speaking skill scores (r =.479 p <.008). In short, the
frequency with which participants reported using L2 words in L1 themselves and how they rated
their speaking skill were two important variables that showed a considerable amount of
relationships with other language background variables.
34
Correlations between background measures and tasks. To determine if performance on
the experimental tasks was related to language background measures, these data were entered
into a correlation analysis. This analysis revealed that if participants scored better on the
vocabulary test of L2, their RTs were faster on both Dutch and English trials, but only in the
semantic decision task (Dutch: r = -.424, p <.02; English: r = -.368, p < .05), suggesting that a
more extensive vocabulary allowed these participants to be more effective in teasing apart Dutch
and English words. Higher L2 vocabulary scores were also associated with faster responses in
the semantic decision task on both congruent (r = -403, p <.03) and incongruent trials (r = -.390,
p <.034) when considering the Position*Button interaction. This relationship was not found in
the language decision task. Finally, self-reported speaking skill showed a negative correlation
with the RT difference between incongruent and congruent trials in the semantic decision task (r
= -.420, p <.022), where higher self-reported speaking skill was associated with a smaller
interference effect in the semantic decision task.
Discussion
The aim of the current study was to investigate executive control in bilingualism by
creating a language decision and semantic decision task that differed only in their instruction as
they involved the exact same stimulus and response set, ensuring that the mental operations that
participants were performing constituted the only difference between the two tasks. In other
words, for each stimulus-response combination (trial), exactly the same response (a left or right
button press) was required for both tasks. Results from these tasks show that despite the use of
35
the exact same stimulus-response set, differences in factor effects and reaction times still ensued
between the tasks due to differences in task demands.
Word semantics and button location were shown to interact in both tasks, with faster RTs
to trials where the semantics of the word matched with the location of the response button (e.g.,
S-Left and the left response button) than RTs to incongruent trials. As expected, we found an
interaction between screen position and button location only in the semantic decision task.
However, the predicted language effect in the language decision task was absent, and this effect
unexpectedly turned up in the semantic decision task, with faster RTs to Dutch than to English
words. In both the language decision and the semantic decision task, we found that RTs to targets
presented on the right of the screen were faster than RTs to targets presented on the left of the
screen. This effect was not predicted, and a possible explanation will be suggested when we
interpret the results. In contrast, the predicted effect of button position was absent in both tasks,
with no faster RTs on the right button despite a right-handed participant group.
In the Simon task, however, we did find an effect of the button position, with
faster responses on the right than on the left response button. Given that all participants were
right-handed, this result was expected. RTs to targets appearing at the screen position congruent
to the location of the response button were faster than RTs to targets at the screen position
incongruent to this location, showing the classic Simon effect in our data. Participants were
hindered by incongruency between screen and button position, as was hypothesized. This effect
was shown to correlate positively with the screen position vs. button location interaction in the
semantic decision task.
Considering the language background measures, participants rated their own production
(writing and speaking) skills as considerably lower than their comprehension (reading and
36
listening) skills). However, the participants’ self-reported speaking skill proved to be important
in the subsequent correlation analyses along with their self-reported frequency of actually using
L2 words in L1 in conversation, showing a considerable number of positive relationships to other
variables, such as the frequency with which the participants’ friends use L2 words in L1 and
their attitude towards the use of L2 words in L1. If someone’s friends use L2 words in L1 more
often, they themselves might use them more as well, and they would perhaps be more positive
towards their use. These results suggested validity of the language background measure. In
relation to the results of the experimental tasks, a larger vocabulary in L2 was found to be
associated with faster RTs on both Dutch and English target words in the semantic decision task,
and on both trials where the position of the target on the screen was congruent or incongruent
with that of the location of the response button.
Interpreting the data
Whilst it was not hypothesized, the observed effect of screen position of the target
in both experimental tasks might stem from the nature of the experimental set-up where
linguistic stimuli were used in a manner that resembles their use in the Simon task. Performance
asymmetries between handedness and visual field presentation have been documented,
suggesting that for linguistic targets presented to right handed people in their right visual field
responses are actually faster than for targets presented in their left visual field. This effect is
thought to stem from the common conception that most functional processing of language is
done by the left hemisphere of the brain in right-handed individuals. This hemisphere also
processes the right visual field. Research has shown faster responses to linguistic stimuli in the
right than the left visual field of right handed participants (Moscovitch, 1976; Calvert, Geddes,
Crow, Norman, & Iversen, 1996). As visual field performance asymmetry is not the main focus
37
of the current study, we shall leave this as a suggested explanation for the position effect.
We found a language (main) effect only in the semantic decision task, which was
counter-intuitive to the design of the experiments, given that we would have expected a language
effect in the language decision task (see figure 2 in the introduction). However, requiring
participants to attend selectively to the semantics of a word does not exclude retrieval of
language node information, and this information could also have an effect on performance. Why
this information did not affect responses in the language decision task, in hindsight, could be
related to the nature of the stimuli. To determine to what language a target word belongs, the first
two letters of each word were actually sufficient. In the case of ‘links’ vs ‘left’ there is a ‘li’ and
‘le’ distinction, that is mirrored in the case of ‘rechts’ and ‘right’ or ‘re’ and ‘ri’. In fact, these
two letters were all participants would have needed to base their language decision on as these
four words were the only targets presented and participants were aware of this. If they based
their performance in the language decision task on this principle, the absence of a language effect
is not unlikely.
In both the language decision and semantic decision tasks there was an interaction
between word semantics and response button location, with smaller RTs to congruent than
incongruent trials. This shows that participants processed the semantics of the target word in
both the language decision and the semantic decision task, and that they were affected by
incongruency between these two factors in both tasks. In the semantic decision task, this effect
was predicted as participants were required to selectively attend to the semantics of the word in
order to follow the instruction of the task. With the largest effect size in the semantic decision
task, the interaction between the semantics of the word and the location of the response button
was also the main effect driving RT differences in this task. In the language decision task,
38
incongruency between the location of the target word and button also caused an interference with
the response. Whilst participants might have based their language decision on superficial
characteristics of the word, its surface orthography would still have activated its deeper meaning,
which consecutively interfered with the required response. If a participant determined the word
‘left’ to be an English word based on ‘le’ and the response button for English was the right
button, the semantic processing of the word ‘left’ could still have interfered with the response on
the right button, as the meaning of the target word was incongruent with the button position. In
this way, there would still be an interaction between semantics and button position in the
language decision task, in the absence of a language effect. Even though participants might have
based their language decision on the beginnings of words, this does not seem to have interfered
with other effects.
The stimuli in the current experiment were very limited in number, but this limited
stimulus set does not seem to have caused any additional problems. In the light of the language
effect in the semantic decision task and the interaction between semantics and button location in
both tasks, it is important to note that no effects were found that suggest a direct interaction
between language membership and word semantics, or between the former two and button
position, in accordance with the BIA+ model. This also shows that the precise word that was
presented did not seem to matter in the semantic decision task: There was no difference between
the different types of incongruency that were listed in table 1 in the introduction. Simply
requiring bilinguals to selectively use the semantic information from direction words was enough
to elicit these effects independent of the specific direction that the words denoted. Because the
semantic decision task required retrieval of semantic information and decision may have been
based on more shallow properties in the language decision task, the semantic decision task may
39
have posed a heavier demand on lexical retrieval, as indicated by the correlations of vocabulary
size and RTs in this task. A larger vocabulary might have been beneficial in the semantic
decision task by a general speed-up in lexical retrieval (Beck, Perfetti, & McKeown, 1982;
Hedden, Lautenschlager, & Park, 2005), while in the language decision task retrieval of more
detailed lexical information was hypothesized to be more of a by-product than a necessity and
language information could be based on shallow stimulus properties.
The most prominent effect in the current study concerned the interaction between the
position of the target on the screen and the location of the response button. We hypothesized that
the selective attention to semantic information in the semantic decision task would lead to
interference outside of the lexicon, at the level of task execution. The results of the current study
support this hypothesis as an interaction between position and button was found only for the
semantic decision task. Furthermore, the hypothesis that this effect should to some degree
resemble the equivalent position vs. button interaction in the Simon task was confirmed, as the
correlational analyses showed that a larger interference effect of screen position and button
location incongruency in the semantic decision task was reliably associated with a larger
interference effect in the Simon task. No such relationship was found for the language decision
task. This finding adds claim to the assumption that the semantic activation in the semantic
decision task gave rise to a spatial interference situation that is similar to that of the classic
Simon effect. Despite the effect of word semantics in the language decision task as evident by
the semantics vs. button location interaction, this semantic activation did not lead to a spatial
interference effect in this task. As mentioned, the specific semantics of the target word were not
found to matter for the spatial interference effect, because no three-way interaction was found
between word semantics, screen position, and button location. Taking this into consideration, the
40
spatial interference effect appears to be due to asking bilinguals to attend to a specific aspect of
words at the task level. If the locus of the effect had solely been the activation of semantic
information in the bilingual lexicon, we should have logically found the same interference effect
in the language decision task, where semantic activation actively influenced the decision process.
Not finding a spatial interference effect in the language decision task despite the evidence of
semantic activation suggests that the spatial interference was caused by a difference between the
tasks located outside the lexicon, in this case the specific task instruction. In short, asking
participants to process the stimuli differently from a cognitive perspective created large
differences in results. This interesting finding suggests a delicate interplay between inhibitory
control within and outside the bilingual lexicon. Performance on language decision and semantic
decision tasks with the addition of spatial information is affected by the interplay between lexical
and spatial information and by task instruction. A change at the level of task schemas was shown
to relate to a change of effects of lexical information.
Having considered the results and their interpretations, we can now turn back to the
model that we proposed in figure 2 of the introduction. This model was a visual representation of
the predictions for both the language decision and the semantic decision task, where language
node information was expected to be the driving factor in the language decision task on the left
of the model, whilst several predictions for the semantic decision task were made on the right. In
light of the previous discussion, we can split the proposed model into two separate models, one
for each task (figure 7).
41
Figure 7. Separate models for the results of (a) the language decision task and (b) the semantic decision task.
For the language decision task, the model has to be simplified considerably as the target
stimulus elicited only an effect of screen position and an interaction between word semantics and
button location. However, the model for the semantic decision task has become slightly more
complicated. Based on the results in this task, the model proposed in the introduction for both
tasks actually applies in its totality to the semantic decision task, with the addition of a main
effect of screen position. This stems from the unexpected effect of language in the semantic
decision task, which was hypothesized but absent in the language decision task. In this way, the
single model proposed for both tasks with a clear separation between the two does not suffice.
For both tasks, aspects of the stimuli that were hypothesized to be of specific interest only to the
response in one task did affect the decision making process in the other task. This finding
suggests that the input information is processed at all levels, and that information that is
irrelevant to the task can still interfere with its execution. Furthermore, the finding that semantic
activation only triggers a spatial interference effect in the semantic decision task suggests that the
locus of the required inhibitory control in the presence of spatial interference resides outside of
the lexicon, at the level of task execution.
The present study has shown that there is a distinction between inhibitory control using
language membership and using spatial information that becomes apparent at the level of task
execution. Here, we have shown how different patterns of interference effects arise when
42
bilinguals perform different cognitive operations on the same stimulus-response set. While the
stimuli and required response type (button press) were the same, bilinguals used different aspects
of those stimuli to come to a decision in the two tasks, and this differential focus on a specific
type of information caused different patterns of interference. Furthermore, finding a spatial
interference effect of screen position vs. button location in the semantic decision task shows that
this spatial information is encoded even when it leads to additional interference with the required
response. This spatial interference effect in the semantic decision task related positively to the
Simon effect, suggesting that the two were similar in nature. Vocabulary size shows a small
negative relationship to performance on the semantic decision task, where an increase in
vocabulary size is related to a decrease in response times.
Instead of comparing bilinguals and monolinguals on a measure of executive control, we
compared multiple measures of linguistic and nonlinguistic control within a group of bilinguals.
Furthermore, we compared the lexical measures directly as each task employed the same
stimulus-response set, only requiring different cognitive operations. The comparison to the
Simon task as a measure of non-lexical inhibitory control was also more direct than in previous
studies, as both our lexical control measures mimicked the presence of spatial information in the
Simon task.
Although we did not use any neuroimaging methods, the use of electroencephalography
(EEG) would be an interesting addition. We have established that instructing participants to
mentally perform a different operation on the same stimuli leads to different results in an RT
task, but the current data provide no information on the time-course of these effects. The use of
EEG could provide such information.
43
Concluding our discussion, models of bilingual inhibitory control will need to consider
the present findings that (a) a target word activates lexical information beyond that required for
the response, (b) this activation might still interfere with the response, and (c) the emphasis on
retrieval of specific lexical information in the bilingual lexicon can influence spatial interference
effects outside of the bilingual lexicon.
References
Beck, I., Perfetti, C., & McKeown, M.G. (1982). Effects of long-term vocabulary instruction on
lexical access and reading comprehension. Journal of Educational Psychology, 74(4),
506-521.
Bialystok, E. (2001). Bilingualism in development: Language, literacy, and cognition. New
York: Cambridge University Press.
44
Bialystok, E. (2006). Effect of bilingualism and computer video game experience on the Simon
task. Canadian Journal of Experimental Psychology, 60, 68–79.
Bialystok, E., Craik, F.I.M., Klein, R., & Viswanathan, M. (2004). Bilingualism, aging, and
cognitive control: Evidence from the Simon task. Psychology and Aging, 19, 290- 303.
Bialystok, E., Craik, F.I.M., & Luk, G. (2008). Cognitive control and lexical access in younger
and older bilinguals. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 34, 859-873.
Bunge, S.A., Dudukovic, N.M., Thomason, M.E., Vaidya, C.J., & Gabrieli, J.D.E. (2002).
Immature frontal lobe contributions to cognitive control in children: Evidence from
fMRI. Neuron, 33, 301–311.
Caramazza, A. (1997). How many levels of processing are there in lexical access? Cognitive
Neuropsychology, 14, 177–208.
Colzato, L.S., Bajo, M.T., van den Wildenberg, W., Paolieri, D., Nieuwenhuis, S., La Heij, W.,
& Hommel, B. (2008). How does bilingualism improve executive control? A comparison
of active and reactive inhibition mechanisms. Journal of Experimental Psychology:
Learning, Memory and Cognition, 34, 302–312.
Costa, A., Miozzo, M. & Caramazza, A. (1999). Lexical selection in bilinguals: Do words in the
bilingual’s two lexicons compete for selection? Journal of Memory and Language, 41(3),
365–397.
Dijkstra, A., de Bruijn, E., Schriefers, H., ten Brinke, S. (2000). More on interlingual homograph
recognition: language intermixing versus explicitness of instruction. Bilingualism:
Language and Cognition, 3, 69-78.
45
Dijkstra, A., Timmermans, M., & Schriefers, H. (2000). On Being Blinded by Your Other
Language: Effects of Task Demands on Interlingual Homograph Recognition. Journal of
Memory and Language, 42, 445-464.
Dijkstra, A., & van Heuven, W. J. B. (1998). The BIA model and bilingual word recognition. In
J. Grainger & A. M. Jacobs (eds.), Localist connectionist approaches to human cognition,
pp.189-225. Mahwah, NJ: Lawrence Erlbaum Associates.
Dijkstra, A., & van Heuven, W.J.B. (2002). The architecture of the bilingual word recognition
system: From identification to decision. Bilingualism: Language and Cognition, 5(3),
175-197.
Emmorey, K., Luk, G., Pyers, J.E., & Bialystok, E. (2008). The source of enhanced cognitive
control in bilinguals. Psychological Science, 19(12), 1201-1205.
Geddes, J., Crow, T.J., Norman, W.D., & Iversen, S.D. (1996). Evidence that manual dominance
and lateralisation of a letter-matching task are decoupled in schizophrenia. Schizophrenia
Research, 18(2), 212-212(1).
Green, D. W. (1986). Control, activation and resource: A framework and a model for the control
of speech in bilinguals. Brain and Language, 27 (2), 210–223.
Green, D. W. (1998). Mental control of the bilingual lexicosemantic system. Bilingualism:
Language and Cognition, 1(2), 67–81.
Hedden, T., Lautenschlager, G., & Park, D. (2005). Contributions of processing ability and
knowledge to verbal memory tasks across the adult life-span. The Quartly Journal of
Experimental Psychology, 58, 169-190.
Kovacs, A. M. (2009) Early bilingualism enhances mechanisms of false-belief reasoning.
Developmental Science, 12 , 48–54.
46
La Heij, W. (2005). Monolingual and bilingual lexical access in speech production: Issues and
models. In Kroll & De Groot (eds.), 289–307.
Levelt,W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech
production. Behavioral and Brain Sciences, 22, 1–75.
Meara, P., & Jones, G. (1990). Eurocentres Vocabulary Size Test. Zurich: Eurocentres Learning
Service.
Meuter, R.F.I., & Allport, A. (1999). Bilingual Language Switching in Naming: Asymmetrical
Costs of Language Selection. Journal of Memory and Language, 40(1), 25-40.
Miyake, A., Freidman, N.P., Emerson, M.J., Witzki, A.H., Howerter, A., & Wager, T.D. (2000).
The unity and diversity of executive functions and their contributions to complex “frontal
lobe” tasks: A latent variable analysis. Cognitive Psychology, 41, 49–100.
Moscovitch, M. (1976). On the representation of language in the right hemisphere of right-
handed people. Brain and Language, 3(1), 47-71.
Perani, D., Abutalebi, J., Paulesu, E., Brambati, S., Scifo, P., Cappa, S.F., & Fazio, F. (2003).
The role of age of acquisition and language use in early, high-proficient bilinguals: An
fMRI study during verbal fluency. Human Brain Mapping, 19(3), 170-182.
Portocarrero, J.S., Burright, R.G., & Donovick, P.J. (2007). Vocabulary and verbal fluency of
bilingual and monolingual college students. Archives of Clinical Neuropsychology, 22,
415-422.
Thomas, M.S.C., & Allport, A. (2002). Language Switching Costs in Bilingual Visual Word
Recognition. Journal of Memory and Language, 43(1), 44-66.
47
Lexical and spatial interference effects in Dutch-English bilinguals and the need for executive control
Appendix a
Table 2. Accuracy percentages by individual participants (two columns of participants) in the language decision and the
semantic decision task.
Language decision task (Task 1) Semantic decision task (Task 2)
Participant Accuracy Participant Accuracy
1 83.9% 1 89.1%
2 86.7% 2 90.1%
3 92.7% 3 91.2%
48
5 87.5% 5 94.3%
6 88.5% 6 95.3%
7 90.6% 7 94.3%
8 84.9% 8 91.2%
9 90.6% 9 95.8%
10 96.4% 10 94.8%
11 93.8% 11 95.3%
12 90.0% 12 97.9%
13 90.6% 13 92.2%
14 91.2% 14 95.8%
15 90.6% 15 95.3%
16 94.3% 16 97.9%
17 92.2% 17 96.4%
18 89.6% 18 93.2%
19 83.3% 19 94.3%
20 90.6% 20 93.8%
21 95.3% 21 99.0%
22 90.6% 22 91.7%
23 84.4% 23 91.7%
24 96.9% 24 98.0%
26 90.6% 26 88.5%
27 85.4% 27 91.2%
28 94.8% 28 92.7%
29 81.3% 29 90.6%
30 94.8% 30 96.4%
31 95.8% 31 96.9%
32 94.8% 32 95.3%
49
top related